With access to large troves of de-identified EHR data shared by its customers, Epic Systems is publishing observational research insights on its Epic Health Research Network (EHRN) website. In a recent interview with Healthcare Innovation, two company executives described EHRN’s goals and some COVID-related work.
Epic describes EHRN as an online journal designed for rapid sharing of knowledge with researchers, healthcare professionals, and learners to help solve medical problems. The company notes that the reports are reviewed internally and externally prior to publication. To expedite information sharing, however, they are published without traditional peer review.
J. Owen Sizemore, Ph.D., data science lead at Epic, said that EHRN “is something we have been thinking about doing for a long time – how powerful it can be if there is some way to more closely link what is happening in medical practice on a large scale and mine information out of that data and provide that to the medical community and speed up that process. Then the pandemic just increased the urgency of it. Very quickly it was an all-hands-on-deck thing to get it up and running.”
Here is an example of the type of research Epic is doing: When the COVID-19 outbreak first emerged, there was concern that patients with urgent conditions would not seek care in emergency departments for fear of exposure to COVID-19. On May 14, 2020, EHRN published a study showing that the weekly incidence of acute myocardial infarctions (AMIs), commonly known as heart attacks, and strokes decreased by 45 percent and 38 percent respectively when compared to historical averages prior to March 13, 2020, the day a national emergency for COVID-19 was issued. Epic then tracked the incidence of these two conditions in the ED for an additional nine weeks and found a dramatic return to near normal levels.
EHR data from ED visits occurring between the weeks of April 7, 2020 and May 31, 2020, show the weekly incidence of AMI has returned to approximately 92 percent of the historical trend prior to March 13, 2020. Similarly, the incidence of stroke has returned to approximately 87 percent of the historical trend.
“With the pandemic, there is not just the issue of what data we can get about COVID treatment to determine what are the best treatments and risk factors,” Sizemore explained, “but we also are interested in how the pandemic is affecting the healthcare system more broadly. We have published a few things about cancer screenings and telehealth, but we wanted to look not just at preventative care, but also emergency care.”
Jeff Trinkle, M.D., who works in clinical informatics at Epic, said the work was partly based on feedback from the Epic community of clinicians who were concerned that people who should be coming into the ED were not. “That community spirit gives us ideas. By looking at this data, we could identify that this does look significant, and that can help customers respond. For instance, here in Madison there has been a lot of education to the public about the precautions hospitals are putting in place. So they can take the next step to get patients comfortable coming back to the ED,” and Epic can help guide patient engagement strategies in other ways, he added.
Here is one way that Epic is helping customers respond to the insights gleaned from this type of data: Groups using Epic reported an 86 to 94 percent reduction in cancer screenings. Reduced schedules have put available appointments at a premium and generated a large backlog of patients waiting to be seen. To help organizations determine which appointments to reschedule first, Epic provides risk-scoring algorithms that look at factors such as age, chronic conditions, obesity, and previous hospitalizations. This information helps them prioritize appointments.
The company quotes Deborah Gash, chief digital officer at Saint Luke’s Health System in Kansas City, Mo., as saying, “Using Epic’s risk-scoring algorithms to re-start services, we can make sure we’re getting the patients who need care most urgently in to see their doctor first.”
Sizemore explained that the data Epic is using is a HIPAA-defined limited data set that does not have individual patient identifiers, but does include relevant medical information. All of the data involved is structured and pulled from discrete fields. “A big reason is that this is a limited data set. In a physician’s note, it is difficult to be 100 percent sure the patient’s name is not in it.”
He also noted that there are limitations to what this type of analysis can accomplish. “Randomized clinical trials are the gold standard and they will continue to be going forward,” he said. “The benefit you get from this database is that it is huge – with so many patients, but the downside is that it is live clinical practice data, so although we are doing analyses on certain things, it is not a replacement for clinical trials. But it can provide insight into places to invest money in a larger clinical trial.”
Trinkl said many of Epic’s customers — both clinicians and leadership — monitor EHRN and see it as a good resource. “I also am aware of patients who have come across it. The articles are designed to be read not just by scientists, but also the public and doctors and any organization that wants to glean some insights. One of our goals is to get researchers saying maybe we should do more research on a topic. I think that would be a really good outcome of what we are doing.”
